Today's chatbots are no longer the dropdown menus of 2015. Modern ones understand natural language: “any room for four tonight at 9?” is read as a real request, not matched by keyword. The most advanced consult documents — the menu, the allergens page, the FAQs — and answer with current information instead of canned lines. It's the leap that made them useful on the floor: they close 70-80% of repetitive questions, opening hours included, without waking anyone. They remain answerers, though: they can tell you the restaurant closes on Mondays, but they don't book Friday's table.
The limit is action. A chatbot can tell you which times are free if someone passes them to it, but it won't go and look: without a connection to the diary it just answers. It doesn't write in the calendar, send the reminder or update the waiting list. That's why the right question isn't “chatbot yes or no” but “what must happen after the answer”: if informing is enough, a chatbot is enough; if a booking must come out of it, you need a system that acts on diary and messages. And the expectation rule holds: promising “I'll book you in” and not doing it is worse than saying straight away that booking is done by hand.
Frequently asked questions
- Can a chatbot really answer like a person?
- Within the limits of a task: repetitive questions and facts to look up, yes, and often well. The moment the conversation turns complex — a complaint, an off-script request, a real doubt — a person is needed, and a good chatbot understands that and steps aside.
- Are a chatbot and an AI agent the same thing?
- No: the chatbot answers, the agent acts. The first stops at words — informs, guides, reassures — the second performs operations: reads the diary, offers free slots, writes the booking and sends the reminder. In practice the chatbot is the voice, the agent is the hand.